EvoPoC
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
๐ก๏ธ EvoPoCA Knowledge-Driven Evolving Agentic System for End-to-End Vulnerability Detection and Exploit Generation end-to-end vulnerability detection โ attack path reasoning โ exploit verification โ executable PoC generation ๐ฌ Demo: Real-World Attack Reproduction (PLTD Token)This demo showcases EvoPoC on a real-world exploited token: PLTD.PLTD was historically attacked and drained on-chainEvoPoC:correctly identifies the vulnerabilityinfers the attack pathgenerates an exploit PoC automaticallyAfter replacing the real contract address and block number, the generated PoC successfully reproduces the attack and extracts 24.4 USDTโโ๐ก Try it yourself! Launch the Web UI with python app.py and experience the interactive analysis interface. See the Web UI section below for detailed interface screenshots and features. Table of ContentsDemo: Real-World Attack ReproductionProject Overview Usage Modes Full Analysis Mode Memory Update Mode (Knowledge Base Evolution) Directory Structure Dependencies How to Run Option 1: CLI Option 2: Web UI Web Interface OverviewOption 3: Use as Library Output Description Notes๐ Project OverviewEvoPoC performs end-to-end smart contract PoC generation, including:CapabilityDescription๐ Protocol Type RecognitionClassifies contract protocol types via Hierarchical Knowledge Graph๐ฏ Root Cause AnalysisUses RAG + LLM to infer vulnerability type and triggering conditionsโ๏ธ Attack Path & Exploit ReasoningGenerates exploit vectors and step-by-step attack plans๐งฎ Symbolic Execution & Constraint SolvingVerifies path feasibility using Z3โ
PoC Generation & Reachability VerificationProduces executable Solidity exploit PoC๐ง Usage ModesFull Analysis ModeEnd-to-end vulnerability discovery and exploit generationUpload contract(s) โ โ vulnerability analysis โ attack path inference โ symbolic verification โ exploit PoC generation โ security report outputMemory Update Mode (Knowledge Base Evolution)Validate existing PoCs and evolve system knowledgeUpload:contract(s)existing exploit PoCSystem will:validate PoC against inferred root causedetect inconsistenciesrefine vulnerability patternsupdate Hierarchical Knowledge Graph and exploit memoryThis enables EvoPoC to:continuously improve its reasoning and exploit generation ability over time๐ Directory Structure EvoPoC/ โโโ main.py # Main orchestration (AnalysisCoordinator / KnowledgeCoordinator) โโโ app.py # Flask web entry โโโ config.py # Global config (API, paths, models) โโโ protocoltypetree.json # Protocol hierarchy โโโ ExploitBehaviortree.json# Exploit behavior tree โ โโโ agents/ # Agent modules โ โโโ protocol_agent.py โ โโโ rootcause_agent.py โ โโโ exploit_agent.py โ โโโ verifier_agent.py โ โโโ rootcauseverify_agent.py โ โโโ poc_validator_agent.py โ โโโ rag/ # Retrieval-Augmented Generation โ โโโ rag_manager.py โ โโโ vector_store.py โ โโโ utils/ โ โโโ file_utils.py โ โโโ json_utils.py โ โโโ z3_utils.py โ โโโ extract/ # Multi-file contract extraction โ โโโ extract_main.py โ โโโ antlr4helper/ โ โโโ rules/ โ โโโ requirements.txt โ โโโ Hierarchical Knowledge Graph/ โ โโโ protocol_types/ โ โโโ rootcause/ โ โโโ ExploitBehavior/ โ โโโ dataset/ โโโ result/ โโโ static/ โโโ templates/ โโโ uploads/ โโโ logs-app/ โโโ demo.mp4 โโโ README.md ๐ฆ DependenciesEnvironmentItemDescription๐ Python3.10+๐ API KeyOpenAI compatible LLM & embeddingInstall pip install -r requirements.txt pip install -r extract/requirements.txt ๐ How to RunOption 1: CLI python main.py Contracts placed under: Vulnerable_Contract/ Option 2: Web UI python app.py Access the web interface at http://127.0.0.1:5000๐ฅ๏ธ Web Interface OverviewMain Analysis DashboardThe main dashboard provides a comprehensive end-to-end analysis workflow: Contract Upload Vulnerability Detection and Root Cause Analysis Knowledge Graph Visualization Interactive force-directed graph visualization showing:Protocol Nodes: Protocol type relationships (blue/green nodes)Root Cause Nodes: Vulnerability root causes (red nodes)Exploit Nodes: Attack steps and exploit paths (orange nodes)Path Validation & Symbolic Execution Comprehensive symbolic execution analysis:Attack Vector Generation: Vulnerability point identification and entry pointsConstraint Collection: Path conditions and critical constraintsZ3 Solving: Constraint solving results with model variablesExecution Trace: Step-by-step execution path analysisProfit & Asset Validation Detailed asset comparison and profit analysis:Attacker Assets: Balance before and after attack executionProfit Calculation: Net profit from the exploitVictim Impact: Contract balance changes before/afterExecution Tracking: Complete execution trace with state changesKnowledge Base Expansion ModeAccess the knowledge base expansion interface at http://127.0.0.1:5000/knowledge_baseThis specialized mode enables:Knowledge Expansion: Automatically expand knowledge base with new vulnerability patternsGraph Updates: Update Hierarchical Knowledge Graph with validated patternsInconsistency Detection: Identify and resolve conflicts between PoC and inferred vulnerabilitiesOption 3: Use as Library from main import analyze_contract_from_file โ result = analyze_contract_from_file("contract.sol") ๐ Output DescriptionFileDescriptionfull_analysis.jsonFull structured reasoning resultsecurity_report.mdHuman-readable security reportexploit_poc.solGenerated exploit PoCโ ๏ธ NotesEnsure FAISS indices exist under Hierarchical Knowledge Graph/Multi-file analysis requires extract/ dependenciesModel names configured in config.py
Fields of Study & Subjects
Blockchain Technology Applications and Security
Software Testing and Debugging Techniques
Security and Verification in Computing
Publication details
- Journal / source
- Figshare
- Publication date
- 1 January 2026
- Language
- EN